Direct Observation of Nitrogen Location in Molecular Beam
Epitaxy Grown Nitrogen-Doped ZnO
P. Fons∗ , H. Tampo∗ , A. V. Kolobov∗ , M. Ohkubo∗ , S. Niki∗ , J. Tominaga∗ , R.
Carboni† and S. Friedrich∗∗
∗
National Institute of Advanced Industrial Science & Technology,Tsukuba Central 4, Higashi 1-1-1, Tsukuba,
Ibaraki, 305-8562, Japan
†
Physics Department and CNISM, University of Bologna, viale C. Berti Pichat 6/2, 40127 Bologna, Italy
∗∗
Advanced Detector Group, L-270 Lawrence Livermore National Laboratory, Livermore, CA 94550, U.S.A.
Abstract. ZnO is a wide band gap, naturally n-type semiconductor with great promise for optoelectronic applications. To
date, however, it has proven difficult to dope p-type, a prerequisite for device fabrication. Nitrogen is widely believed to be
one of the most promising dopant candidates, however, experimental results to date have been inconsistent; recent theoretical
formation energy calculations have indicated that Nitrogen preferentially incorporates into the ZnO lattice in the form of a N−
2
molecule at an O-site when a Nitrogen plasma source is used, leading to compensation rather than p-type doping. We show by
a combination of X-ray absorption spectroscopy at the N K-edge of plasma-assisted molecular beam epitaxy grown ZnO and
ab-initio simulations that in as-grown material, Nitrogen incorporates substitutionally on an O-site where it is expected to act
as an acceptor. We have also observed the distinctive formation of molecular nitrogen bubbles upon rapid thermal annealing.
These results suggest that effective p-type doping of ZnO with N may only be possible for metastable low-temperature growth
processes.
Keywords: ZnO, Nitrogen, doping, XAFS
PACS: 61.10Ht, 71.15MB, 68.55.Ln, 66.30.Jt
INTRODUCTION
Zinc Oxide (ZnO) has been the subject of renewed research over the past several years in large part to its similarities to GaN. Both crystallize in the wurtzite structure
(P63 mc) and have comparable lattice constants (GaN: a=
0.3189, c = 0.5185, ZnO: a= 0.3250, c=0.5207 Å) and
bandgaps (GaN:3.36 eV, ZnO: 3.35 eV). ZnO, however,
has several advantages not shared by GaN or its alloys
including the facts that its constituents are plentiful and
inexpensive, bulk substrates are available and it can be
allowed with Mg to form the pseudobinary Mgx Zn1−x O
alloy allowing the growth of single-phase alloys with
bandgaps up to 4.0 eV.
Much of the utility of ZnO in optoelectronic or spintronic applications lies requires the fabrication of p-type
material, a property that has proven to be elusive. Reports
to date on the fabrication of p-type ZnO have not been
reproduced by other groups and the lack of a broadly
accepted technique for p-type doping is limiting the applications of ZnO. The combination of very low activation energy for n-type dopants, the existence of several
dominant native defect and hydrogen related donor levels [1] located within 0.01-0.05 eV of the conduction
band edge combined with the anticipated relatively high
activation of p-type dopants leads to ambiguity in electrical results in the study of p-type doping due to strong
carrier compensation effects. Nitrogen is widely considered to be the most promising p-type dopant due to its
similar size to oxygen, resistance to forming AX centers [2], and its successful use in other II-VI compounds
such as ZnSe [3].
In situations such as the above, theoretical results can
often provide insight. Recently, density functional theory (DFT) has been applied to determine the formation energy of various types of compensating defects
for nitrogen-doped ZnO in the presence of a plasma
source[4]. By using a value for the N chemical potential that took into account the presence of a mixture
of ground and excited state N2 in the doping flux, it
was shown that the calculated N maximum solubility increased over eight orders of magnitude with the use of a
plasma N source. Concomitant with the use of a plasma
N source, it was also found that the N-related defect with
the lowest formation energy was a N2 defect at an Osite. The conclusion drawn was that compensation effects from the double donor N2 defect would dominate
and lead to electron-based conduction. Since these calculations depend strongly on the atomic chemical potentials and the fermi level and would seem to rule out
the efficacy using Nitrogen as a p-type dopant, confirmation by experiment is desirable. To this end, we have applied soft x-ray absorption at the K-edge of N (403 eV)
and first-principles real-space scattering simulations to
EXPERIMENT
Although epitaxial ZnO has been grown by a variety of
techniques including RF sputtering[5], MO-chemical vapor deposition[6], and laser-ablation[7] among others,
we focus here on radical source-molecular beam epitaxy
(RS-MBE)[8]. In RS-MBE, Zn flux is provided by an
elemental (7N) effusion cell while O and N fluxes originate from RF-radical (gas) cells. Details of the growth
procedure have been described previously [9]. The film
growth temperature was fixed at 450◦ C for all samples;
the sample thickness was approximately 1 µm. Samples
with a variety of N2 flow rates with differing II/VI ratios
were grown. For the as-grown samples, calibrated SIMS
depth profiling measurements indicated a nearly constant
N signal with average N concentrations approaching 1%
(9 × 1020 cm−3 ). This value may be compared with the
electron densities . 1 × 1017 cm3 and . 1012 cm3 for an
undoped ZnO film and the hydrothermal ZnO substrates,
respectively.
After growth, all samples were divided into two
pieces: one half was retained as an as-grown standard,
while the other half underwent rapid thermal annealing (RTA) to investigate thermal stability. All RTA samples were heated to 800◦ C for three minutes. It was
noted that as-grown samples appeared transparent, but
yellow in color, while the same samples after RTA treatment were perfectly transparent. Hall effect electrical
measurements made using the Van der Pauw geometry revealed for a typical N-doped sample grown on
the Zn face of a ZnO wafer, the initial carrier concentration was . 5 × 1015 cm3 (µ ∼ 10 cm2 /Vs), rising to
∼ 6 × 1017 cm−3 (µ ∼ 70 cm2 /Vs) after RTA processing.
The N-doped ZnO samples were measured at two locations the ELETTRA synchrotron in Trieste, Italy and
the Advanced Light Source at Lawrence Berkeley National Laboratory. The ELETTRA measurements were
carried out at beamline ALOISA. ALOISA is an undulator beamline with a photon flux of approximately
1011 photons/sec. The beam was linearly polarized and
was incident upon the sample at approximately 50◦ degrees. X-ray fluorescence radiation was detected using a windowless single element Ge solid-state detector (SSD). The use of photon counting allowed measurement of samples from the high 1019 cm−3 concentration range. Due to the close proximity of the O K-edge
(∼ 540 eV), only near-edge (. 100 eV) measurements at
the N K-edge (∼ 409 eV) were carried out. Crystals in
the wurtzite structure have two distinct XANES patterns,
one for polarization along the [001] axis and another
for polarization in-plane. For non-polarization dependent
measurements, an incident angle of approximately 50◦
was used resulting in a weighted sum of the two possible polarization spectra; the sample geometry was taken
into account for all simulations. Polarization-dependent
measurements were taken using a novel multielement superconducting detector at beamline 4.0.2 (photon flux
∼ 1012 photons/second) of the Advanced Light Source
at Lawrence Berkeley National Laboratory [10].
RESULTS AND DISCUSSION
XAFS measurements on an as-grown N-doped ZnO sample along with simulated results are shown in Fig.1. Data
was interpreted using a two step process. Ground-state
DFT calculations were used to relax the ZnO lattice
around various N lattice positions and then these relaxed
coordinates were used as input to a real-space, full multiple scattering simulation for comparison with experiment.
experiment
a
Absorption (arb. units)
directly observe the location of N in the ZnO lattice both
in as-grown and thermal annealed states.
b
c
d
e
400
420
440
Photon Energy (eV)
f
460
FIGURE 1. Experimental (a) and simulated real space multiple scattering XANES simulations for various DFT relaxed N
local environments including (b) N on an O site, (c) N2 on an
O site, (d) N in a tetrahedral interstice, (e) N in a Zn site, and
(f) N in an octahedral interstice. The energy step was 0.5 eV.
In the first step, an impurity N atom was placed at
a high symmetry position in the ZnO lattice and the
surrounding atoms were relaxed to their ground state
positions using Broyden minimization. Potentials were
calculated self-consistently using a conjugate-gradient
algorithm[11]. The norm-conserving pseudopotentials in
CONCLUSION
The DFT calculations in general showed significant lattice relaxation for Nitrogen locations other than on the
O-site resulting in peak broadening and a significant discrepancy from experiment. The simulated spectra for
Nitrogen incorporating on an O site showed excellent
agreement with experiment. After annealing, the samples
showed only a single peak. A more detailed examination
of both the as-grown and annealed samples was carried
out at beamline 4.0.2 at the ALS; this data can be seen in
Fig.2. The annealed spectra are an excellent match to the
characteristic spectra of the molecular vibration levels of
a N2 molecule. In addition, as the spectra show no polarization dependence, it is strong suggested that the Nitrogen contained in the films has transform to N2 , presumably in the form of bubbles. This behavior is also consistent with the change in sample color from yellow to
transparent as N moves from electronically active sites to
electronically inactive bubbles. In contrast, no such fine
structure is visible in the as-grown film seen in Fig.2(e).
X-ray Absorption
the Troullier-Martins scheme used for the calculation included the 3d10 4s2 (Zn), 2s2 2p4 (O), and 2s2 2p3 (N)
valence electrons as well as non-local core corrections.
The generalized gradient approximation (GGA) as expressed by the Perdew-Burke-Ernzerhof exchange formulation was used [12]. Calculations were carried out
on 32 atom supercells using a plane wave cutoff energy
of 50 Hartree. The final atom positions as expressed by
the relaxed supercell where embedded into a DFT relaxed ZnO spherical cluster and used as input to the
ab-initio real space full multiple-scattering (FMS) code
FEFF 8 for the calculation of the simulated x-ray absorption spectra[13]. Sudies of cluster size using FEFF 8
demonstrated that convergence was reached for cluster
sizes greater than 200 atoms, hence all results presented
here are for clusers that size. Fig.1 shows the results of
these calculations while an interpretation of these results
is provided in the next section.
400.5
401.0
401.5
402.0
Photon Energy (eV)
FIGURE 2. Polarization measurements made for (a) a
molecular N2 gas standard and on a N-doped ZnO sample after
annealing for E (b) along the c-axis, (c) at 45◦ to the c-axis, (d)
in the a-b plane. (e) shows the equivalent range for an as-grown
sample. The energy step was 40 meV
ment of Energy by Lawrence Livermore National Laboratory under contract No. W-7405-ENG-48.
REFERENCES
1.
2.
3.
4.
5.
6.
ACKNOWLEDGMENTS
7.
The DFT results were obtained through the use of
the ABINIT code, a common project of the Université Catholique de Louvain, Corning Incorporated, and
other contributors (URL http://www.abinit.org). We are
also grateful to the ALOISA staff for help with the experiment. Measurements at ELETTRA were supported
by the Synchrotron Radiation Committee of Istituto
Nazionale per la Fisica della Materia. S.F. acknowledges
NSF, DOE OBER and NA-22 funding. Part of this work
was performed under the auspices of the U.S. Depart-
8.
9.
10.
11.
12.
13.
C. Van de Walle, Phys. Rev. Lett. 85, 1012–1015 (2000).
C. H. Park, S. B. Zhang, and S.-H. Wei, Phys. Rev. B 66,
073202 (2002).
K. Ohkawa, T. Karasawa, and T. Mitsuyu, J. Cryst.
Growth 111, 797–801 (1991).
E. Lee, Y. Kim, Y. Jin, and K. Chang, Physica B 308,
912–915 (2001).
Y. Igasaki, and H. Saito, J. Appl. Phys. 69, 2190–2195
(1991).
Y. Liu, C. Gorla, S. Liang, N. Emanetoglu, Y. Lu, H. Shen,
and M. Wraback, Journal of Electronic Materials 29,
69–74 (2000).
S. Choopun, R. Vispute, W. Noch, A. Balsamo, R. Sharma,
T. Venkatesan, A. Iliadis, and D. Look, Appl. Phys. Lett.
75, 3947–3949 (1999).
P. Fons, K. Iwata, A. Yamada, K. Matsubara, and S. Niki,
J. Crystal Growth 201-202, 627 (1999).
P. Fons, K. Iwata, A. Yamada, K. Matsubara, K. Nakahara,
T. Tanabe, H. Takasu, and S. Niki, Appl. Phys. Lett. 12,
1801–1803 (2000).
S. Friedrich, T. Funk, O. Drury, S. Labov, and S. Cramer,
Rev. Sci. Inst. 73, 1629–1631 (2002).
X. Gonze, Phys. Rev. B 54, 4383 (1996).
J. Perdew, K. Burke, and M. Ernzerhof, Phys. Rev. Lett.
77, 3865–3868 (1996).
A. Ankudinov, and J. Rehr, Phys. Rev. B 62, 2437 (2000).